This is a revision of our multi-site collaborative Ancillary Study to the STAR*D Project, in which we propose to study biomarkers that may help guide treatment by prospectively identifying patients with treatment resistant depression (TRD). Pilot data from both randomized clinical trials and naturalistic treatment show that a quantitative electroencephalographic (QEEG) measure, cordance, is associated with response and nonresponse to pharmacotherapy; changes in prefrontal activity early in treatment are predictive of later response. Preliminary data suggest that a biomarker model using the cordance indicator can identify (a) patients who will be resistant to treatment prescribed in Level 2 of the STAR*D protocol, and (b) patients who will have an initial but non-sustained, "placebo-like" response. Such prospective identification would allow physicians to undertake more sophisticated regimens earlier and attain improved clinical outcome. The revisions in this proposal include expansion of the rationale for using the cordance biomarker, additional detail about the STAR*D protocol, additional pilot data, and subject flow projections that have been refined to reflect recent STAR*D enrollment figures. The three specific aims of this project are: (1) to evaluate the use of neurophysiologic biomarkers in an effectiveness treatment trial setting for acceptability and cost effectiveness; (2) to ascertain the validity of cordance indicators in prospectively identifying depressed patients who will be resistant to their Level 2 treatment choices; and (3) to examine the use of these indicators in identifying patients who will not have a sustained response to treatment. We will test specific hypotheses: (1) assessment with QEEG will be acceptable and offer cost effective guidance in an effectiveness treatment trial; (2) acute treatment response in Level 2 to specific switching or augmentation strategies will be predicted by changes in prefrontal cordance in the first two weeks of Level 2 treatment; and (3) sustained treatment response will be predicted by changes in prefrontal values after two weeks of treatment. 72 subjects with depression will be recruited when they enter treatment in Level 2 of the protocol at either of two STAR*D sites, 36 each at UCLA's Neuropsychiatric Institute and Harvard's Massachusetts General Hospital. QEEG data will be recorded at the start of Level 2 and after two additional weeks. Treating clinicians and subjects will be blinded to physiologic data, and outcomes will be assessed using the instruments integral to the STAR*D protocol. Subjects will be followed-up at 3 and 6 months after completing Level 2 to assess whether responses are sustained. The test of proportion and t-tests will be used to evaluate acceptability and cost effectiveness. Prediction of acute response will be tested with chi square and linear regression models. Independent examination of the data at the two sites will be used to assess the generalizability of the cordance biomarker method. The relationship of our biomarkers to sustained response will be tested by chi square analyses. Secondary analyses will use general linear models and growth mixture modeling to examine how early cordance changes relate to patient and illness factors and to functional outcomes. [unreadable] [unreadable]